MCPcopy Index your code

hub / github.com/congwei1230/MoChaBench / functions

Functions46 in github.com/congwei1230/MoChaBench

⨍ Functions 46◇ Types & classes 10
↓ 6 callersFunctionmean
(vals)
eval-lipsync/script/run_syncnet_pipeline_on_mocha_generation_on_mocha_bench.py:124
↓ 6 callersFunctionmean
(vals)
eval-lipsync/script/run_syncnet_pipeline_on_your_own_model_results.py:125
↓ 3 callersMethodinference
( self, video_path: str, # We do not extract audio from video_path! audio_path: str,
eval-lipsync/script/syncnet_pipeline.py:182
↓ 2 callersMethodforward
(self, x)
eval-lipsync/script/detectors/s3fd/nets.py:21
↓ 1 callersMethod__init__
( self, num_classes=2, top_k=750, nms_thresh=0.3, conf_thresh=0.05,
eval-lipsync/script/detectors/s3fd/box_utils.py:139
↓ 1 callersMethod__init__
(self, n_channels, scale)
eval-lipsync/script/detectors/s3fd/nets.py:10
↓ 1 callersMethod_crop
(self, track, frames, audio_wav, base)
eval-lipsync/script/syncnet_pipeline.py:135
↓ 1 callersMethod_iou
(a, b)
eval-lipsync/script/syncnet_pipeline.py:94
↓ 1 callersMethod_load_s3fd
(self, path: str)
eval-lipsync/script/syncnet_pipeline.py:78
↓ 1 callersMethod_load_syncnet
(self, path: str)
eval-lipsync/script/syncnet_pipeline.py:85
↓ 1 callersMethod_track
(self, dets)
eval-lipsync/script/syncnet_pipeline.py:102
↓ 1 callersFunctioncalc_pdist
(feat1, feat2, vshift=10)
eval-lipsync/script/SyncNetInstance.py:19
↓ 1 callersFunctioncompute_qwk_list
Compute Quadratic Weighted Kappa between human and GPT scores (4 items). Returns NaN if undefined.
eval-viescore/compute_alignment.py:17
↓ 1 callersFunctiondecode
Decode locations from predictions using priors to undo the encoding we did for offset regression at train time. Args: loc (tensor): lo
eval-lipsync/script/detectors/s3fd/box_utils.py:41
↓ 1 callersMethoddetect_faces
Same detection code as before, but we no longer load the model here.
eval-lipsync/script/detectors/s3fd/__init__.py:28
↓ 1 callersFunctionencode_image_from_array
(image_array)
eval-viescore/eval_gpt_viescore.py:32
↓ 1 callersMethodevaluate
(self, opt)
eval-lipsync/script/SyncNetInstance.py:51
↓ 1 callersFunctionextract_frames
(video_path, num_frames=NUM_FRAMES)
eval-viescore/eval_gpt_viescore.py:36
↓ 1 callersMethodforward_aud
(self, x)
eval-lipsync/script/SyncNetModel.py:83
↓ 1 callersMethodforward_lip
(self, x)
eval-lipsync/script/SyncNetModel.py:89
↓ 1 callersMethodforward_lipfeat
(self, x)
eval-lipsync/script/SyncNetModel.py:95
↓ 1 callersMethodfrom_dict
(cls, d: Dict[str, Any])
eval-lipsync/script/syncnet_pipeline.py:52
↓ 1 callersFunctionload_prompt
(metric, prompt_idx=None)
eval-viescore/eval_gpt_viescore.py:60
↓ 1 callersFunctionmain
()
eval-lipsync/script/run_syncnet_pipeline_on_mocha_generation_on_mocha_bench.py:89
↓ 1 callersFunctionmain
()
eval-lipsync/script/run_syncnet_pipeline_on_your_own_model_results.py:90
↓ 1 callersFunctionnms
Apply non-maximum suppression at test time to avoid detecting too many overlapping bounding boxes for a given object. Args: boxes: (te
eval-lipsync/script/detectors/s3fd/box_utils.py:66
↓ 1 callersFunctionnms_
Courtesy of Ross Girshick [https://github.com/rbgirshick/py-faster-rcnn/blob/master/lib/nms/py_cpu_nms.py]
eval-lipsync/script/detectors/s3fd/box_utils.py:7
↓ 1 callersFunctionparse_output
(text)
eval-viescore/eval_gpt_viescore.py:52
↓ 1 callersMethodreset_parameters
(self)
eval-lipsync/script/detectors/s3fd/nets.py:18
↓ 1 callersFunctionrun_sample
(row)
eval-lipsync/script/run_syncnet_pipeline_on_mocha_generation_on_mocha_bench.py:54
↓ 1 callersFunctionrun_sample
(row)
eval-lipsync/script/run_syncnet_pipeline_on_your_own_model_results.py:54
↓ 1 callersFunctionspearman_footrule_list
Compute Spearman's footrule distance between human and GPT rankings (4 items).
eval-viescore/compute_alignment.py:8
Method__init__
( self, cfg: Union[PipelineConfig, Dict[str, Any], None] = None, *, device: st
eval-lipsync/script/syncnet_pipeline.py:60
Method__init__
(self, num_layers_in_fc_layers=1024)
eval-lipsync/script/SyncNetModel.py:17
Method__init__
( self, net: torch.nn.Module, device: str = "cuda", dropout: float = 0,
eval-lipsync/script/SyncNetInstance.py:40
Method__init__
We now accept an *already-initialized* S3FDNet as `net`, instead of loading weights here.
eval-lipsync/script/detectors/s3fd/__init__.py:15
Method__init__
( self, input_size, feature_maps, variance=[0.1, 0.2], min_sizes=[16,
eval-lipsync/script/detectors/s3fd/box_utils.py:191
Method__init__
(self, device="cuda")
eval-lipsync/script/detectors/s3fd/nets.py:29
Functionevaluate_task
(task)
eval-viescore/eval_gpt_viescore.py:71
Methodextract_feature
(self, opt, videofile)
eval-lipsync/script/SyncNetInstance.py:156
Methodforward
(self, loc_data, conf_data, prior_data)
eval-lipsync/script/detectors/s3fd/box_utils.py:155
Methodforward
(self)
eval-lipsync/script/detectors/s3fd/box_utils.py:211
Methodforward
(self, x)
eval-lipsync/script/detectors/s3fd/nets.py:111
Functionload
(filename)
eval-lipsync/script/SyncNetModel.py:11
MethodloadParameters
(self, path)
eval-lipsync/script/SyncNetInstance.py:206
Functionsave
(model, filename)
eval-lipsync/script/SyncNetModel.py:5